Phage T7 helicase unwinds double-stranded DNA (dsDNA) by encircling one strand while excluding the complementary strand from its central channel. When T7 helicase translocates on single-stranded DNA (ssDNA), it has kilobase processivity; yet, it is unable to processively unwind linear dsDNA, even 60 base-pairs long. Particularly, the GC-rich dsDNAs are unwound with lower amplitudes under single-turnover conditions. Here, we provide evidence that T7 helicase switches from ssDNA to dsDNA during DNA unwinding. The switching propensity is higher when dsDNA is GC-rich or when the 3′-overhang of forked DNA is <15 bases. Once helicase encircles dsDNA, it travels along dsDNA and dissociates from the end of linear DNA without strand separation, which explains the low unwinding amplitude of these substrates. Trapping the displaced strand with ssDNA binding protein or changing its composition to morpholino oligomer that does not interact with helicase increases the unwinding amplitude. We conclude that the displaced strand must be continuously excluded and kept away from the central channel for processive DNA unwinding. The finding that T7 helicase can switch from ssDNA to dsDNA binding mode during unwinding provides new insights into ways of limiting DNA unwinding and triggering fork regression when stalled forks need to be restarted.